Comparison of Cold and Warm Extremes in the Mediterranean Based on E-OBS and ERA5-Land Data Sets

Accurate characterization of temperature extremes is essential for climate risk assessment, particularly in regions highly sensitive to climate change like the Mediterranean Basin. This study compares near-surface temperatures from E-OBS and ERA5-Land, focusing on warm (>90th percentile) and cold (<10th percentile) extremes. Our findings indicate substantial temporal agreement across most of the domain, with event overlaps in the period 1991–2020 (last decade) up to 74% (86%) for warm extremes and 64% (80%) for cold extremes, though certain sub-regional sectors emerge as structural outliers with weaker coherence. For warm extremes, a systematic bias is prevalent: E-OBS is generally 1.1°–1.3°C warmer than ERA5-Land at the 90th percentile, though this divergence narrows at higher intensities. Distribution analyses reveal a pronounced negative skewness in certain areas, reflecting localized overestimations by the reanalysis. Regarding decadal warming trends, ERA5-Land generally exhibits steeper rates for warm extremes, reaching up to 0.86°C/decade compared to 0.68°C/decade in E-OBS, despite lower absolute temperatures. For cold extremes, both data sets capture coherent warming signals across most areas, maintaining high correlations (0.94–0.98), whereas the outlier sector shows a clear trend inconsistency with a weak or negative warming signal in the reanalysis. Comparative spatial analyses confirm that complex topography amplifies data dispersion and intensifies the reanalysis positive differences in mountainous terrains, as shown by wider probability density functions in elevated areas. Crucially, regional station density modulates trend coherence, as high-density areas show uniform warming shifts across all percentiles. Overall, both data sets confirm their ability to reproduce seasonality and exhibit a regional warming signal.